Design of an Automated-Counting System of Cell Micronuclei in Micrographs

摘要

We developed and tested a system for the automatic analysis of cell images in order to identify and count micronuclei configurations in digital images from a microscope. The presence of micronuclei has been used as an indicator of DNA damage, and a fast automated analysis may be suitable for early cancer detection, for example. We describe in this work the image-processing protocol and system comprising: acquisition, color normalization, contrast enhancement, color-background removing, color segmentation, mathematical-morphology filtering and restoration, morphometry, feature extraction and analysis, and counting. Among the morphological features used for discriminating micronuclei configurations, we tested compactness, area ratios and separation of selected features (ideally, a nucleus and one or more micronuclei). Among the mathematical-morphology techniques, we used for instance a modified watershed segmentation algorithm for separating touching features. We present our results on several images, the evaluation and performance of each step, and the main problems we solved by using several techniques from mathematical morphology and color-image processing.